Defect generation factor estimation device, defect generation factor estimation method, defect generation factor estimation model learning method, operation condition determination method, and steel product manufacturing method

Abstract

This defect generation factor estimation device is provided with: a defect generation depth calculation means for calculating the depth position of a surface defect in a product in a slab stage; a defect generation rate calculation means for calculating a defect generation rate for each slab depth; an inclusion adhesion rate calculation means for calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting on the basis of the defect generation rate for each slab depth; and a defect generation factor estimation means for estimating a defect generation factor by selecting a portion of the manufacturing conditions using an estimation model learned in advance using the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data and estimating the inclusion adhesion rate when the selected manufacturing conditions are changed in stages.

Description

Technical Field

[0001] The present invention relates to a defect generation factor estimation device, a defect generation factor estimation method, a learning method for a defect generation factor estimation model, an operation condition determination method, and a manufacturing method of a steel product. Background Art

[0002] Patent Document 1 discloses an auxiliary system that assists in a cause estimation operation when an operator performs an operation to estimate the cause of a quality abnormality. In the method proposed in Patent Document 1, focusing on the correlation between operation variables and quality in past manufacturing performance data, operation variables with a strong correlation with quality are extracted, and a time series graph of the operation variables is displayed to the operator.

[0003] Prior Art Documents

[0004] Patent Documents

[0005] Patent Document 1: Japanese Patent No. 6116445 Gazette Summary of the Invention

[0006] Problems to be Solved by the Invention

[0007] In the method proposed in Patent Document 1, in order to determine the cause of a quality abnormality, cause estimation is performed based on product quality data and manufacturing condition data, focusing on the relationship based on the correlation between quality and manufacturing conditions. However, simply focusing on the correlation between operation factors and quality cannot estimate at which stage of the solidification process of molten steel inclusions that become defect seeds adhere and become surface defects in the final product. In order to perform an operation for eliminating surface defects in the final product, it is extremely important to grasp the relationship between the solidification process of molten steel and the adhesion of inclusions.

[0008] The present invention has been completed in view of the above circumstances, and an object thereof is to provide a defect generation factor estimation device, a defect generation factor estimation method, a learning method for a defect generation factor estimation model, an operation condition determination method, and a manufacturing method of a steel product that can estimate defect generation factors based on the relationship between the solidification process of molten steel and the adhesion of inclusions.

[0009] Means for Solving the Problems

[0010] In order to solve the above problems and achieve the object, the defect generation factor estimation device according to the present invention includes: a defect generation depth calculation unit that calculates the depth position of a surface defect in a product at the slab stage; a defect generation rate calculation unit that calculates the defect generation rate for each slab depth based on the aforementioned depth position; an inclusion adhesion rate calculation unit that calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect generation rate for each slab depth; and a defect generation factor estimation unit that uses a pre-learned estimation model with the manufacturing conditions of the aforementioned product as input data and the aforementioned inclusion adhesion rate as output data, selects a part of the manufacturing conditions, and estimates the inclusion adhesion rate when the selected manufacturing conditions are changed step by step, thereby estimating the defect generation factor.

[0011] In the defect generation factor estimation device according to the present invention, in the above invention, the aforementioned estimation model is a model pre-learned with manufacturing conditions including the distance from the solidification start position and operating conditions in continuous casting as input data and the inclusion adhesion rate for each distance from the solidification start position as output data.

[0012] In the defect generation factor estimation device according to the present invention, in the above invention, the aforementioned defect generation depth calculation unit calculates the aforementioned depth position by at least considering the grinding amount of the slab grinding machine during slab finishing, the scale peeling amount during hot rolling, and the reduction amount of the plate thickness due to pickling.

[0013] In the defect generation factor estimation device according to the present invention, in the above invention, the aforementioned defect generation factor estimation unit displays the relationship between one or more manufacturing conditions and the aforementioned inclusion adhesion rate as one-dimensional data or multi-dimensional data.

[0014] In the defect generation factor estimation device according to the present invention, in the above invention, the aforementioned inclusion adhesion rate is the probability of bubbles in molten steel or inclusions as minute solids adhering during the casting process.

[0015] In the defect generation factor estimation device according to the present invention, in the above invention, the aforementioned manufacturing conditions include any one or more of the tundish gas blowing flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity.

[0016] In order to solve the above problems and achieve the object, the method for estimating defect generation factors according to the present invention includes: a step of a defect generation depth calculation unit provided in a computer for calculating the depth position in the slab stage of a surface defect in a product; a step of a defect generation rate calculation unit provided in the computer for calculating the defect generation rate for each slab depth based on the depth position; a step of an inclusion adhesion rate calculation unit provided in the computer for calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect generation rate for each slab depth; and a step of a defect generation factor estimation unit provided in the computer for using a previously learned estimation model that takes the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data, selecting a part of the manufacturing conditions, and estimating the inclusion adhesion rate when the selected manufacturing conditions are changed step by step, thereby estimating the defect generation factors.

[0017] In order to solve the above problems and achieve the object, the learning method for an estimation model of defect generation factors according to the present invention includes: a step of a defect generation depth calculation unit provided in a computer for calculating the depth position in the slab stage of a surface defect in a product; a step of a defect generation rate calculation unit provided in the computer for calculating the defect generation rate for each slab depth based on the depth position; a step of an inclusion adhesion rate calculation unit provided in the computer for calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect generation rate for each slab depth; and a step of a model learning unit provided in the computer for learning an estimation model for estimating the inclusion adhesion rate by taking the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data.

[0018] In order to solve the above problems and achieve the object, the operating condition determination method according to the present invention determines the operating conditions based on the defect generation factors estimated by the above-described defect generation factor estimation method.

[0019] In order to solve the above problems and achieve the object, the manufacturing method of a steel product according to the present invention manufactures a steel product based on the operating conditions determined by the above-described operating condition determination method.

[0020] In order to solve the above problems and achieve the object, the manufacturing method of a steel product according to the present invention estimates the inclusion adhesion rate for each distance from the solidification start position during slab casting by the above-described defect generation factor estimation method and based on the operating conditions after casting, and changes the operating conditions of subsequent processes based on the estimated inclusion adhesion rate, thereby manufacturing a steel product.

[0021] Effects of the Invention

[0022] In the apparatus and method for estimating defect generation factors according to the present invention, focusing on the reproducibility of the solidification process of molten steel and the attachment of inclusions, a estimation model is used to estimate at which stage of the solidification process of molten steel inclusions that become defect seeds attach and appear as surface defects of the product based on multiple past operation performance data. Thus, the relationship between the solidification process of molten steel and the attachment of inclusions can be grasped, and therefore operations for manufacturing products without surface defects can be considered. In addition, according to the learning method of the defect generation factor estimation model according to the present invention, an estimation model for estimating defect generation factors can be constructed based on the relationship between the solidification process of molten steel and the attachment of inclusions. In addition, according to the operation condition determination method according to the present invention, operations for manufacturing products without surface defects can be implemented. In addition, according to the manufacturing method of steel products according to the present invention, steel products without surface defects can be manufactured. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Figure 1 FIG. is a diagram showing a schematic configuration of a defect generation factor estimation apparatus according to an embodiment of the present invention.

[0024] Figure 2 Figure 2 FIG. is a flowchart showing steps of a defect generation factor estimation method executed by the defect generation factor estimation apparatus according to an embodiment of the present invention.

[0025] Figure 3 Figure 3 FIG. is a diagram showing an example of a manufacturing process of a stainless steel product (steel product).

[0026] Figure 4 Figure 4 This is an example of the present invention, and FIG. is a diagram showing an example of manufacturing performance data read from a database in the data reading step of the defect generation factor estimation method.

[0027] Figure 5 Figure 5 This is an example of the present invention, and FIG. is a superimposed graph showing the total number of defects for each slab depth calculated in the defect generation rate calculation step of the defect generation factor estimation method.

[0028] Figure 6 Figure 6 This is an example of the present invention, and FIG. is a line graph showing the defect generation rate for each slab depth calculated in the defect generation rate calculation step of the defect generation factor estimation method.

[0029] Figure 7 Figure 7 This is an example of the present invention, and FIG. is a schematic diagram showing the solidification of molten steel and the flow of molten steel in a mold.​​​​​​​​​​​​​​

[0030] Figure 8 Figure 8 is an embodiment of the present invention and is a line graph showing the inclusion adhesion rate for each meniscus distance calculated in the inclusion adhesion rate calculation step of the defect generation factor estimation method.

[0031] Figure 9 Figure 9 is an embodiment of the present invention and is a histogram showing the model accuracy of the estimation model constructed in the model learning step of the defect generation factor estimation method.

[0032] Figure 10 Figure 10 is an embodiment of the present invention and is a contour map showing the relationship between the distance from the meniscus, the tundish gas flow rate, and the inclusion adhesion rate estimated in the defect generation factor estimation step of the defect generation factor estimation method.

[0033] Figure 11 Figure 11 is an embodiment of the present invention and is a graph showing the relationship between the distance from the meniscus and the inclusion adhesion rate estimated in the defect generation factor estimation step of the defect generation factor estimation method.

[0034] Figure 12 Figure 12 is an embodiment of the present invention and is a graph showing the relationship between the slag basicity and the inclusion adhesion rate estimated in the defect generation factor estimation step of the defect generation factor estimation method. Detailed Embodiments

[0035] Regarding the defect generation factor estimation device, the defect generation factor estimation method, the learning method of the defect generation factor estimation model, the operation condition determination method, and the manufacturing method of steel products according to the embodiments of the present invention, description will be made with reference to the accompanying drawings. Hereinafter, an example in which the present invention is applied to a steelmaking process for manufacturing thick plate steel will be described. However, the present invention is not limited to the following embodiments, and the constituent elements in the following embodiments also include elements that can be easily replaced or substantially the same elements by those skilled in the art.

[0036] (Defect Generation Factor Estimation Device)

[0037] Regarding the defect generation factor estimation device according to the embodiment, reference is made to Figure 1 for description. The information processing device 10 for implementing the defect generation factor estimation device includes an arithmetic processing unit 101, a storage unit (ROM: Read Only Memory) 103, a temporary storage unit (RAM: Random Access Memory) 104, a bus wiring 105, and a database (DB) 120.

[0038] ​​​​​​​​​​The arithmetic processing unit 101 is implemented using an electronic circuit such as a CPU (Central Processing Unit), etc., and executes the defect generation factor estimation program 102 in the storage unit 103, performing various arithmetic processes required for the defect generation factor estimation process. The arithmetic processing unit 101, as a functional module that functions by executing the defect generation factor estimation program 102, includes a data reading unit 106, a defect generation depth calculation unit 107, a defect generation rate calculation unit 108, an inclusion adhesion rate calculation unit 109, a model learning unit 110, and a defect generation factor estimation unit 111.

[0039] The data reading unit 106 executes the data reading step described later. In addition, the defect generation depth calculation unit 107 executes the defect generation depth calculation step described later. In addition, the defect generation rate calculation unit 108 executes the defect generation rate calculation step described later. In addition, the inclusion adhesion rate calculation unit 109 executes the inclusion adhesion rate calculation step described later. In addition, the model learning unit 110 executes the model learning step described later. In addition, the defect generation factor estimation unit 111 executes the defect generation factor estimation step described later. The detailed content of each step will be described later (see Figure 2 ).

[0040] The defect generation factor estimation program 102 is stored in the storage unit 103. Past manufacturing performance data is stored in the database 120. It should be noted that, for example, as Figure 1 shown, a display device 20 for displaying the estimation result of the defect generation factor and an input device 30 for receiving input from an operator may also be connected to the information processing device 10.

[0041] (Defect Generation Factor Estimation Method)

[0042] Regarding the defect generation factor estimation method executed by the defect generation factor estimation device according to the embodiment, refer to Figure 2 for description. The defect generation factor estimation method includes a data reading step, a defect generation depth calculation step, a defect generation rate calculation step, an inclusion adhesion rate calculation step, a model learning step, and a defect generation factor estimation step. It should be noted that in the defect generation factor estimation method according to the embodiment, the steps up to the data reading step to the model learning step, and the defect generation factor estimation step may also be implemented at other times. That is, in the defect generation factor estimation method according to the embodiment, a estimation model may also be pre-constructed through the data reading step to the model learning step, and the defect generation factor estimation step using the estimation model may be implemented at a different time.

[0043] Hereinafter, each step of the defect generation factor estimation method according to the embodiment will be described. It should be noted that the defect generation factor estimation method starts processing, for example, when the target material is manufactured and the defect generation distribution is measured by the defect generation factor estimation device.

[0044] <Data Reading Step>

[0045] In the data reading step, the data reading unit 106 reads the manufacturing performance data of continuous casting stored in the database 120 (step S1). This manufacturing performance data is data obtained by summarizing, for each product (final product), the manufacturing conditions associated with the thickness change of the slab (semi-finished product) in a plurality of past-produced coils (steel strips), the manufacturing conditions associated with the generation of surface defects, and whether there are defects on the surface of the coil. In addition, the manufacturing performance data is composed of, for example, matrix data with the products manufactured in the past represented in the row direction and the manufacturing conditions and inspection results (presence or absence of defects) represented in the column direction (see below Figure 4 ).

[0046] <Defect Generation Depth Calculation Step>

[0047] In the defect generation depth calculation step, the defect generation depth calculation unit 107 calculates the depth position (defect generation depth position) in the slab stage of the surface defect in the product based on the manufacturing conditions read in the data reading step (step S2). The "defect generation depth position" represents the depth position in the slab where the inclusions that appear as surface defects in the final product are attached.

[0048] In the defect generation depth calculation step, specifically, the reduction amount in the thickness direction of the product in each process is calculated based on the manufacturing conditions associated with the thickness change of the slab, and thus it is calculated at which depth the surface defect in the product is in the slab stage. It should be noted that examples of the "manufacturing conditions associated with the thickness change of the slab" include rolling reduction, heating temperature, acid concentration and pickling speed of pickling, and the number of passes of the slab and coil grinders, etc., which are manufacturing conditions for removing scale, grinding, and dissolving the surface of the product.

[0049] In the defect generation depth calculation step, for example, at least the grinding amount of the slab grinder in slab finishing, the scale removal amount in hot rolling, and the plate thickness reduction amount (pickling dissolution amount) due to pickling are considered to calculate the above-mentioned depth position. It should be noted that in the defect generation depth calculation step, the defect generation depth in the slab stage is calculated for all the products read in the data reading step.

[0050] <Defect Generation Rate Calculation Step>

[0051] In the defect generation rate calculation step, the defect generation rate calculation unit 108 calculates the defect generation rate for each specified slab depth based on the depth position in the slab stage (step S3). In the defect generation rate calculation step, for each slab for which the defect generation depth has been calculated in the defect generation depth calculation step, the number of defective products and the total number of all products are summed at a specified slab depth (specified depth interval, e.g., 0.5 mm). Then, the defect generation rate (defect generation ratio) for each specified slab depth is calculated by dividing the number of defective products by the total number of all products. It should be noted that in the defect generation rate calculation step, the defect generation rate for each specified slab depth is calculated for all products read in the data reading step. Here, in the above description, the specified depth interval in the depth direction is set to be constant, but as long as the distribution of the defect generation rate in the slab depth direction is known, it does not necessarily have to be a constant interval.

[0052] <Inclusion adhesion rate calculation step>

[0053] In the inclusion adhesion rate calculation step, the inclusion adhesion rate calculation unit 109 converts the slab depth determined in the defect generation rate calculation step into the distance from the start of solidification at the time of slab casting that becomes the shell thickness equivalent to that slab depth (hereinafter referred to as "distance from the meniscus") (step S4). Thereby, the inclusion adhesion rate for each distance from the meniscus is calculated based on the defect generation rate for each slab depth.

[0054] Here, the inclusion adhesion rate calculated in the inclusion adhesion rate calculation step is the probability that bubbles in the molten steel or inclusions as minute solids adhere during the casting process. It should be noted that in the inclusion adhesion rate calculation step, the inclusion adhesion rate for each distance from the meniscus is calculated for all products read in the data reading step.

[0055] <Model learning step>

[0056] In the model learning step, the model learning unit 110 learns a prediction model, a defect generation factor prediction model for predicting the inclusion adhesion rate (hereinafter referred to as "prediction model"), using the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data (step S5).

[0057] In the model learning step, the learning of the prediction model is performed based on the operation performance data read in the data reading step and the inclusion adhesion rate for each distance from the meniscus calculated in the inclusion adhesion rate calculation step. More specifically, in the model learning step, the manufacturing conditions including the distance from the meniscus and the operation conditions in continuous casting are used as input data, and the inclusion adhesion rate for each distance from the meniscus is used as output data for the learning of the prediction model.

[0058] The learning of the estimation model in the model learning step can be performed by regression methods such as random forest, linear regression, support vector machine, neural network, etc. However, as long as it is a method for estimating the target variable based on the explanatory variables, the learning of the estimation model can also be performed by methods other than the above regression methods.

[0059] <Defect Generation Factor Estimation Step>

[0060] In the defect generation factor estimation step, the defect generation factor estimation unit 111 inputs the value of the explanatory variable of the operating condition to be estimated and the distance from the meniscus into the estimation model constructed in the model learning step. Thereby, the inclusion adhesion rate within the distance from the meniscus under this operating condition is calculated (step S6).

[0061] In the defect generation factor estimation step, specifically, using an estimation model that has been pre-learned with the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data, a part of the manufacturing conditions is selected and the inclusion adhesion rate when the selected manufacturing conditions are changed step by step is estimated. Thereby, the defect generation factor is estimated. As the manufacturing conditions input to the estimation model, for example, it includes any one or more of the tundish gas blowing flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity.

[0062] In addition, in the defect generation factor estimation step, for example, the relationship between one or more manufacturing conditions and the inclusion adhesion rate is displayed as one-dimensional data or multi-dimensional data (see Figures 10 to 12 ).

[0063] (Example)

[0064] Regarding the embodiments of the present invention, reference is made to Figures 3 to 12 for description. Hereinafter, an example of applying the defect generation factor estimation method according to the present invention to the manufacturing of stainless steel products will be described.

[0065] Figure 3 An example of the manufacturing process of stainless steel products is shown. In the manufacturing of stainless steel products, after refining, casting, slab finishing, hot rolling, pickling and annealing 1 (first pickling and annealing), cold rolling, pickling and annealing 2 (second pickling and annealing), defect determination is performed in the slit inspection process. Then, it is shipped as a coiled product. Hereinafter, it will be described according to each step of the defect generation factor estimation method according to the present invention (see Figure 2 ).

[0066] <Data Reading Step>

[0067] First, in the data reading step, read as Figure 4The shown manufacturing performance data. The manufacturing performance data in this figure is matrix data where the rolled products manufactured in the past are represented in the row direction, and the manufacturing conditions and inspection results of each process are represented in the column direction. Additionally, in the manufacturing performance data of this figure, the coil No. is recorded in the first row, and the name of the item is recorded in the first column. The number of samples (number of rows) of this manufacturing performance data is 2300 coils.

[0068] <Defect generation depth calculation step>

[0069] In the defect generation depth calculation step, using the following formulas (1) to (5), the reduction amount in the thickness direction of the slab caused by peeling, grinding, and dissolution in each process is calculated from the coil data in the second row of the Figure 4 manufacturing performance data. Then, the reduced thickness amount in terms of slab from the completion of casting to becoming a product is calculated using the following formula (6). Thus, the defect generation depth in the slab stage of the coil is calculated.

[0070] (A) Slab grinding amount in slab finishing = ●● [mm] / per pass × number of passes of the grinding machine ··· (1)

[0071] (B) Surface scale peeling amount during heating in hot rolling = ●● [mm] ··· (2)

[0072] (C) Pickling dissolution amount in pickling and annealing 1 = ●● [g / m 2 × specific gravity 7.5 × ●● [g / m 3 / width of the slab [m] × (reference pickling speed [mpm] / pickling speed [mpm]) ··· (3)

[0073] (D) Coil grinding amount in pickling and annealing 1 = ●● [μm] / per pass × number of passes of the grinding machine ··· (4)

[0074] (E) Pickling dissolution amount in pickling and annealing 2 = ●● [g / m 2 × specific gravity 7.5 × ●● [g / m 3 / width of the slab [m] ··· (5)

[0075] (F) Reduction amount in slab thickness from the completion of casting to becoming a product = (A) + (B) + (C) × slab thickness / pickling and annealing slab thickness + (D) × slab thickness / pickling and annealing slab thickness + (E) × slab thickness / product thickness ··· (6)

[0076] Additionally, in the defect generation depth calculation step, for all 2300 coils from the third row onwards of the Figure 4 manufacturing performance data, the above calculations of formulas (1) to (6) are performed to calculate the defect generation depth in the slab stage of each coil.

[0077] <Defect generation rate calculation step>

[0078] In the defect generation rate calculation step, based on the data of the defect generation depth in the slab stage of each coil, the defect generation rate is calculated at intervals of 0.5 mm of the slab depth. Figure 5 It is a graph showing the total result of the defect generation depth represented by a superimposed graph obtained by superimposing the number of defect-free slabs (= the number of coils. When multiple coils are generated from one slab, it is converted to the number of slabs) and the number of defective slabs. In this graph, the horizontal axis represents the slab depth (at intervals of 0.5 mm), and the vertical axis represents the number of slabs. Additionally, Figure 6 is based on Figure 5 The calculated defect generation rate, with the horizontal axis representing the slab depth (at intervals of 0.5 mm) and the vertical axis representing the defect generation rate.

[0079] <Inclusion attachment rate calculation step>

[0080] In the inclusion attachment rate calculation step, by the following formula (7), the Figure 6 shown slab depth is converted into the distance from the meniscus (solidification start point) of the mold of the continuous casting machine in the drawing direction.

[0081] Slab depth (shell thickness) [m] = solidification constant [ms -1 / 2 × (distance from the meniscus [m] / casting speed [ms -1 ^(1 / 2) ··· (7)

[0082] Regarding the meaning of the above formula (7), refer to Figure 7 for explanation. This figure shows a schematic diagram of the solidification of molten steel in the mold and the flow of molten steel. As shown in this figure, in the mold, inclusions that become defect seeds adhere to the shell, but the shell thickness when inclusions adhere exactly exposes the defects on the surface as defects in the product stage when the slab is reduced through each process. Therefore, the distance from the meniscus and the shell thickness in the above formula (7) have the same meaning as the slab depth. Additionally, the "defect generation rate within a specified slab depth" can be regarded as the "inclusion attachment rate within a specified distance from the meniscus".

[0083] Figure 8 Shows the result obtained by converting the slab depth on the horizontal axis of Figure 6 into the distance from the meniscus. The horizontal axis represents the distance from the meniscus, and the vertical axis represents the inclusion attachment rate. It should be noted that along with the conversion of the slab depth into the distance from the meniscus, in Figure 8 the Figure 6 defect generation rate on the vertical axis is expressed as the inclusion attachment rate. As Figure 8As shown, it can be seen that in the vicinity of the meniscus, at positions 30 to 40 mm, 70 to 80 mm, and 90 to 120 mm from the meniscus respectively, inclusions are likely to adhere. That is, in this figure, it is indicated that the molten steel flow transports inclusions towards the above-mentioned positions.

[0084] In the model learning step, the explanatory variables (input data) were set as the meniscus distance, tundish blown gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity, and the target variable was set as the inclusion adhesion rate. A random forest of the machine learning method was used to learn the estimation model. Figure 9 It is a histogram showing the model accuracy of the learned estimation model. The horizontal axis represents the inclusion adhesion rate estimated by the estimation model, and the vertical axis represents the number of verification samples (frequency). In addition, in this figure, the histogram showing the coils presumed to be defect-free (without inclusion adhesion) is represented by dotted shading, and the histogram showing the coils presumed to be defective (with inclusion adhesion) is represented by diagonal shading.

[0085] In Figure 9 it, the peak of the histogram showing the coils presumed to be defect-free and the peak of the histogram showing the coils presumed to be defective represented by diagonal shading are clearly separated left and right. Therefore, it can be considered that the learned estimation model can accurately estimate whether the coils are defective.

[0086] <Defect generation factor estimation step>

[0087] In the defect generation factor estimation step, for the above-mentioned learned estimation model, the values of the casting speed, casting width, casting thickness, slab length, and slag basicity among the explanatory variables were fixed at the average values of actual operations and input. In addition, the two explanatory variables of the meniscus distance and the tundish blown gas flow rate among the explanatory variables were changed and input. In this way, in the defect generation factor estimation step, by fixing a part of the input data and changing other input data step by step and inputting them into the estimation model, the inclusion adhesion rate was estimated.

[0088] Figure 10 It is a contour map showing the relationship between the distance from the meniscus, the tundish blown gas flow rate, and the inclusion adhesion rate estimated by the above method. In this figure, the vertical axis represents the distance from the meniscus, the horizontal axis represents the tundish blown gas flow rate, and the dot density represents the inclusion adhesion rate. In this figure, the distance from the meniscus is farther as it goes up. In addition, the tundish blown gas flow rate is larger as it goes to the right. In addition, the inclusion adhesion rate is higher as the dot density is larger.

[0089] By referring to Figure 10, it can be read that the inclusion adhesion rate is high near the meniscus, and the more the gas flow rate blown into the tundish is increased, the lower the inclusion adhesion rate can be. Therefore, the operator can understand the relationship between the solidification process of molten steel and the adhesion of inclusions by referring to the results of this figure, and can get inspiration for considering the optimal molten steel flow (for suppressing the amount of inclusion adhesion, that is, surface defects of the product).

[0090] Here, in Figure 10 In the present invention, the relationship between multiple explanatory variables (distance from the meniscus, gas flow rate blown into the tundish) and the inclusion attachment rate is estimated, but the relationship between one explanatory variable and the inclusion attachment rate can also be estimated.

[0091] For example, Figure 11 This is a graph showing the relationship between the distance from the meniscus and the inclusion adhesion rate among the explanatory variables. In this graph, the horizontal axis represents the distance from the meniscus and the vertical axis represents the inclusion adhesion rate. In addition, the solid line, the dotted line, and the single-point dashed line represent the inclusion adhesion rate when the tundish injection gas flow rate (TD injection gas flow rate) is 4L / min, 24L / min, and 58L / min, respectively. In this way, the defect adhesion rate at each distance from the meniscus (= slab depth) can also be estimated based on the operating conditions after casting.

[0092] in addition, Figure 12 This is a graph showing the relationship between slag basicity and inclusion adhesion rate among the explanatory variables. In this graph, the horizontal axis represents slag basicity and the vertical axis represents inclusion adhesion rate. In this way, the defect adhesion rate of each slag basicity can also be estimated based on the operating conditions after casting.

[0093] In addition, the description so far assumes that the relationship between the solidification process of molten steel and the adhesion of inclusions is understood based on the estimation result of the defect generation factor estimation step, and the operating conditions for eliminating the surface defects of the product are determined, so that the effect is exerted in the next operation. The above method can also be replaced by the following method, for example, in the defect generation factor estimation step, the inclusion adhesion rate during slab casting is estimated based on the operating conditions after casting, and the operating conditions of the subsequent process are changed based on the estimation result. That is, when the product is manufactured, the inclusion adhesion rate can be estimated at the stage of the midway process, and the operating conditions of the subsequent process can be determined (changed) based on the estimation result.

[0094] In this case, for example, as mentioned above Figure 11As shown, the defect adhesion rate for each distance from the meniscus ( = slab depth) is estimated based on the operating conditions after casting. Then, based on the estimated defect adhesion rate, the slab depth amount where the defect adhesion rate is greater than the threshold value determined from the perspective of quality assurance is read (for example, when the defect adhesion rate is 20%, it is within 2 mm from the slab surface layer). Then, the grinding amount of the slab grinding machine in the slab finishing process as a subsequent process is determined in such a way that only the read slab depth amount is finally cut from the slab surface layer.

[0095] That is, the amount obtained by subtracting the predicted values of the surface scale peeling amount in hot rolling and the pickling dissolution amount in pickling and annealing from the amount of the slab to be cut is determined as the grinding amount of the slab grinding machine. It should be noted that when the determined grinding amount of the slab grinding machine exceeds the pre-set maximum grinding amount of the grinding machine (for example, the maximum grinding amount determined from the perspective of repair efficiency), for example, the setting of reducing the pickling speed is changed.

[0096] In addition, for example, as shown above Figure 12 The defect adhesion rate for each slag basicity in the refining process is estimated. Here, the slag basicity is restricted from the operating perspective other than inclusion generation such as the manufacturing cost of the slag. Therefore, for example, when the slag basicity is restricted to 1.25 - 2.00%, from Figure 12 The value of the slag basicity with the minimum inclusion adhesion rate of 1.6% is read. Then, the read slag basicity (1.6%) is determined as the set value of the slag basicity in the refining process.

[0097] In this way, when manufacturing a product under pre-set operating conditions, the probability of defects occurring at which depth in the slab is estimated, and based on the estimation result, the operating conditions are changed to appropriate ones, thereby enabling the manufacture of a product without surface defects.

[0098] In the defect generation factor estimation device and defect generation factor estimation method according to the above-described embodiment, attention is paid to the reproducibility of the solidification process of the molten steel and the adhesion of inclusions. And a estimation model is used that estimates at which stage of the solidification process of the molten steel inclusions that become defect seeds adhere and appear as surface defects of the product based on multiple past operation performance data. Thereby, the relationship between the solidification process of the molten steel and the adhesion of inclusions can be grasped, and thus the operations for manufacturing a product without surface defects can be considered.

[0099] In addition, according to the learning method of the defect generation factor estimation model according to the embodiment, an estimation model for estimating defect generation factors can be constructed based on the relationship between the solidification process of the molten steel and the adhesion of inclusions.

[0100] In addition, according to the method for determining operating conditions according to the embodiment, the operating conditions can be determined based on the defect generation factors deduced by the above-described defect generation factor deduction method. Thus, operations for manufacturing products without surface defects can be implemented.

[0101] In addition, according to the manufacturing method of steel products according to the embodiment, steel products can be manufactured based on the operating conditions determined by the above-described operating condition determination method. Thus, steel products without surface defects can be manufactured.

[0102] In addition, according to the manufacturing method of steel products according to the embodiment, the inclusion adhesion rate at each distance from the solidification start position during slab casting can be deduced by the above-described defect generation factor deduction method and based on the operating conditions after casting. Then, the operating conditions of subsequent processes can be changed based on the deduced inclusion adhesion rate, thereby manufacturing steel products. Thus, steel products without surface defects can be manufactured.

[0103] As described above, embodiments applying the invention completed by the inventors of the present application have been described, but the present invention is not limited to the descriptions and drawings that are part of the disclosure of the present invention constituting the present embodiment. That is, all other embodiments, examples, and techniques applied by those skilled in the art based on the present embodiment are included in the scope of the present invention.

[0104] Description of Reference Numerals

[0105] 10 Information processing device

[0106] 101 Arithmetic processing unit

[0107] 102 Defect generation factor deduction program

[0108] 103 Storage unit (ROM)

[0109] 104 Temporary storage unit (RAM)

[0110] 105 Bus wiring

[0111] 106 Data reading unit

[0112] 107 Defect generation depth calculation unit

[0113] 108 Defect generation rate calculation unit

[0114] 109 Inclusion adhesion rate calculation unit

[0115] 110 Model learning unit

[0116] 111 Defect generation factor deduction unit

[0117] 120 Database (DB)

[0118] 20 Display device

[0119] 30 Input device

Claims

1. Defect generation factor estimation device, comprising: A defect generation depth calculation unit that calculates the depth position of a surface defect in a product at the slab stage; A defect generation rate calculation unit that calculates the defect generation rate for each slab depth based on the depth position; An inclusion adhesion rate calculation unit that calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect generation rate for each slab depth; And A defect generation factor estimation unit that uses a pre-learned estimation model with the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data, selects a part of the manufacturing conditions, and estimates the inclusion adhesion rate when the selected manufacturing conditions are changed step by step, thereby estimating the defect generation factors.

2. The defect generation factor estimation device according to claim 1, wherein, The estimation model is a model pre-learned with manufacturing conditions including the distance from the solidification start position and operating conditions in continuous casting as input data and the inclusion adhesion rate for each distance from the solidification start position as output data.

3. The defect generation factor estimation device according to claim 1, wherein, The defect generation depth calculation unit calculates the depth position by at least considering the grinding amount of the slab grinding machine in slab finishing, the scale peeling amount in hot rolling, and the plate thickness reduction amount due to pickling.

4. The defect generation factor estimation device according to claim 1, wherein, The defect generation factor estimation unit displays the relationship between one or more manufacturing conditions and the inclusion adhesion rate as one-dimensional data or multi-dimensional data.

5. The defect generation factor estimation device according to claim 1, wherein, The inclusion adhesion rate is the probability that bubbles or inclusions as minute solids in molten steel adhere during casting.

6. The defect generation factor estimation device according to claim 1, wherein, The manufacturing conditions include any one or more of the tundish gas blowing flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity.

7. Defect generation factor estimation method, which comprises: A step in which a defect generation depth calculation unit provided in a computer calculates the depth position of a surface defect in a product at the slab stage; A step in which a defect generation rate calculation unit provided in the computer calculates the defect generation rate for each slab depth based on the depth position; A step in which an inclusion adhesion rate calculation unit provided in the computer calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect generation rate for each slab depth; and A step in which a defect generation factor estimation unit provided in the computer uses a pre-learned estimation model with the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data, selects a part of the manufacturing conditions, and estimates the inclusion adhesion rate when the selected manufacturing conditions are changed step by step, thereby estimating the defect generation factors.

8. Learning method of defect generation factor estimation model, which comprises: A step in which a defect generation depth calculation unit provided in a computer calculates the depth position of a surface defect in a product at the slab stage; The defect occurrence rate calculation unit provided in the computer calculates the defect occurrence rate for each slab depth based on the depth position; The inclusion adhesion rate calculation unit provided in the computer calculates the inclusion adhesion rate at each distance from the solidification start position during slab casting based on the defect generation rate at each slab depth; and The model learning unit provided in the computer learns an estimation model for estimating the inclusion adhesion rate using the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data.

9. An operating condition determining method for determining an operating condition based on the defect causing factor estimated by the defect causing factor estimating method according to claim 7.

10. A method for producing a steel product, the method comprising producing the steel product based on the operating conditions determined by the operating condition determination method according to claim 9.

11. A method for manufacturing a steel product, comprising estimating the inclusion adhesion rate at each distance from the solidification start position during slab casting based on the operating conditions after casting by using the defect generation factor estimation method according to claim 7, and changing the operating conditions of subsequent steps based on the estimated inclusion adhesion rate, thereby manufacturing the steel product.

Images